IT IS WELL-KNOWN that vector fields on the disc, or more generally on a planar surface, do not exhibit complicated global phenomena: by the Poincare-Bendixson theorem, every recurrent orbit is either a singularity or a simple closed curve. The situation is completely different for line fields. In particular, it was shown in [6] that it is possible for an orbit of a line field to lose itself in a labyrinth, i.e. to enter a disc and stay inside for ever without either going to a singularity or spiralling towards a compact cycle. Line fields in the disc and the annulus were studied and classified in [6], under the assumption that the singularities are thorns and tripods. It was also asked in [6] to what extent line fields with nontrivial recurrence can be differentiable. We intend in this paper first to partially answer this question, then to extend the classification in [6] to the case when singularities are thorns and saddles with any number of prongs. First we discuss differentiability. There are several natural definitions of C’differentiability for singular line fields (even for orientable ones). First one can require the line field to be C’ in the complement of the singularities; it was pointed out in [6] that being C” in this sense does not restrict the dynamics of the foliation. Another notion of differentiability imposes strong C’ local models near the singularities (see 92). Assuming C2-differentiability in this sense, we prove that the Poincart-Bendixson theorem continues to hold; that is, every orbit entering a disc and staying inside has to go to a singularity or spiral towards a compact cycle. The proof is inspired in part by Denjoy’s theorem about C2 vector fields on the torus. This is in sharp contrast with the fact due to Cherry [l], that on the torus minus a disc there are vector fields (Cm in the above sense) where every orbit coming in from the boundary stays inside forever and does not spiral towards a compact cycle. A very interesting question is whether the Poincare-Bendixson theorem continues to hold for line fields with hyperbolic-type singularities; by this we mean that the natural holonomy maps obtained by following a separatrix through a singularity are of the same form as for hyperbolic vector fields. In $3 and $4 of this paper, we extend the results of [6] to line fields with thorns and n-prong saddles. For the disc, the situation is not essentially different from the case of thorns and tripods. However, for the annulus, the study is quite different and more complicated. In both cases, we obtain a topological structure theorem for the foliation. In several instances the proofs we give simplify arguments in [6].
Read full abstract